Gold rush of .si names

The registry's own counters tell the story faster than any analysis can. Here is who is buying, how little of it is end-user demand, and the detail nobody is discussing: this rush is being prospected by AI.

The counter

These are the live figures published on the .si registry's own front page. We recommend reading them before anything else written about this market, including ours.

Set that against the baseline. The .si namespace held roughly 186,000 domains in June 2026. It now holds 315,921. The namespace has grown by about 70% in a matter of weeks — and the 24-hour figure says the rush is still accelerating, not fading.

Who is booking these domains

Not Slovenians, mostly. Hostinger reported roughly 4,300 registrations on 30 September and about 5,400 on 1 October, with a majority of buyers from the United States and India.

The registrar concentration is equally revealing. Dynadot alone holds 33,572 .si domains — 12.8% of measured registrations. That is not a pattern of distributed local demand. That is a pattern of portfolio buyers using the registrars with the best bulk tooling.

How many are end users?

On the available evidence, close to none that can be measured. Three independent signals point the same way.

Read those together. Half the inventory was for sale within days of being bought. Only three names in a hundred even gesture at the theme that supposedly caused the rush. And the platform people actually use to build something saw nothing at all.

There is a matching signal in the aftermarket. One survey found .com domains pairing brand names with "superintelligence" listed at $20,000 each — and no evidence that any had actually sold and changed hands. Asking prices are not transactions.

The strange part: this rush is AI-prospected

Here is the detail we have not seen discussed anywhere, and it is the most genuinely novel thing about this event.

Previous land rushes were worked by hand. Investors kept word lists, ran them through registrar bulk-check forms, and traded spreadsheets. The constraint was human throughput, which is why the good names took weeks to disappear rather than hours.

This one is different. In the course of our own research for this series, we swept all 17,576 possible three-letter combinations under .si and found roughly 79% already registered. Of 262 widely used three-letter business and technology acronyms — CRM, ERP, API, SDK, KPI, ROI, CEO, LLM, RAG, GPU, SEO, IPO — only five remained unregistered. We generated and screened several thousand four-letter candidates and found the dictionary words almost entirely gone while invented strings survived in bulk.

That pattern is a fingerprint. A namespace picked over that evenly, that quickly, across that many categories, is not being worked by people with word lists. It is being worked by models generating candidate sets and scripts checking them in parallel. The tooling that used to belong to a handful of large portfolios is now available to anyone who can write a prompt.

This matters commercially in two directions. It compresses the window in which a rush is exploitable from weeks to roughly 48 hours — which is exactly what the registry's own September figures show. And it means the names left over after an AI-prospected rush are not the ones that were overlooked. They are the ones that were evaluated and rejected, which is a very different proposition for anyone arriving late.

What happens at month twelve

The question that decides whether any of this was real is the renewal.

In April 2026 Verisign disclosed first-year renewal rates for .com and .net for the first time in years: the mid-40% range for first-year registrations, against the mid-80% range for everything renewing a second time or later. Over half of all new registrations never reach their second year — on the most established extension on earth. For .xyz, roughly 93% of registrations never see a second year at all.

The .si rush cohort faces that test in September 2027. At $12 a year, many will renew out of sheer inertia, which will flatter the number without creating a market. We would watch the development rate instead: how many of these names are serving a website twelve months from now. On the .tv precedent, where fewer than 30% of high-value sales were ever developed, the honest expectation is low.